Back in the late 1950s, George A. Philbrick Researches would cheerfully sell you a $22 K2-W op amp that would swing ±50 V as its rated output (per my old column*), and for $33, a K2-X that would swing ±100 V of output. But there were problems with the K2-X.... After the assemblers built up a group of 50 of these K2-Xs (with all the resistors and capacitors wired to an Octal Plug, and a couple of nine-pin sockets), a test technician would sit down with 50 12AX7A tubes and 50 6AN8s and try to get them all working.
Typically, the 12AX7As gave no problems, but the 6AN8s would not all work. After the first 30 amplifiers were matched up and tested out, the last 20 amplifiers would not meet specs, no matter which tubes were mated with which amplifiers. So, we could not ship 50 op amps until the test technician cheated and borrowed some tubes from the next batch of tubes.
Well, that gets our first 50 out, but how about the next batch? We can't borrow those tubes forever. At the end of the week, there might be a few dozen op amps that would not work with any tube and several dozen tubes that would not work with any amplifier. H'mmm.
Ahem. What exactly was wrong with the circuit? Apparently, the tolerances piled up, and the Plate of the pentode would not swing 230 V p-p (see the figure). In those days, it was—apparently—easier to buy a few hundred more 6AN8s and keep refitting the tubes and the amplifiers together.
Who designed the K2-X? Did Roger ("Tim") Noble design it? Maybe. Roger was a very bright and talented guy who could solve problems even without a college degree. He could make any K2-X work. But he could not make every K2-X work.
Roger did a redesign of the K2-X, to be called the K2-XA. It used a more consistent triode-pentode, the 6BR8A. When he had cooked up a good circuit, I remember Roger called everybody over for a demonstration. He showed us the new, improved circuit, with poor gain and poor swing. Then he added a resistor, and the gain magically popped into spec.
We were all impressed that the new K2-XA had a chance to work better. But I was a bit skeptical. I asked Tim which resistor he had changed. He said he had increased the resistance in the screen circuit. So the distortion cancelled out, and the gain improved. I thought about it and went away, not saying much.
Months later, we realized that the K2-XA had the same kind of "selective fit" problems. The 6BR8As did not provide any magical 100% yield, or 95 %, or even 85% or 75%.... The basic problem was that the pentode could not satisfactorily swing 220 V p-p with a 300-V power supply. If the pentode turned OFF, the plate still would not go high enough. If the pentode turned on very hard, its plate voltage would still not go low enough. The resistor tolerances would build up, the neon lamps' operating voltages would have tolerance errors, the 6BR8A's tolerances would build up, and nobody could make all the darned things work.
I am sure Tim tried various trim schemes to put a trim resistor across some of the resistors to bring the circuit into operation. But part of the problem was that a 4.7-MΩ resistor would have to be shunted by a trim resistor of 62 or 100 MΩ, and in those days, Allen Bradley did not make those high values. They did, later, in the 1965 era. Check out the next issue for the "final solution" to the problem.
Comments invited! [email protected] —or:
Mail Stop D2597A, National Semiconductor
P.O. Box 58090, Santa Clara, CA 95052-8090
*Electronic Design, Feb. 3, 2003, p. 22.